Field of the Invention
The present invention relates to techniques to output sound from sound output devices such as robots and smartphones. Particularly, the invention relates to techniques concerning sound output devices that are capable of accepting messages through a microphone, and outputting messages through a speaker.
Description of the Related Art
Techniques to output sound from sound output devices such as robots and smartphones are known. Sound output devices that are capable of accepting messages through a microphone, and outputting messages through a speaker are also proposed.
For example, JP-A-2002-091482 (Patent Literature 1) discloses an emotion detection method, an emotion detection device, and a storage medium. The emotion detection method described in this publication is a method for detecting subject's emotions, whereby the sound level, the tempo indicative of the rate of occurrence of sound, and the intonation indicative of the pattern of sound level changes in each word are detected from input sound signals, and the amounts of change of the detected sound level, tempo, and intonation are determined to generate signals indicative of the states of the emotions of at least anger, sadness, and joy based on the amounts of change determined.
JP-A-2007-296169 (Patent Literature 2) discloses a mood determining device, and a program. The mood determining device of this publication determines subject's physiological excitement (hereinafter, “biological excitement”) by using biological information. The device also determines subject's sound excitement (hereinafter, “vocal excitement”) by using subject's sound signals. These two types of excitement (biological excitement and vocal excitement) are compared to infer a subject's mood.
Japanese Patent No. 4851447 (Patent Literature 3) discloses a sound analysis device, a sound analysis method, and a sound analysis program for detecting pitch frequency. The sound analysis device of this publication includes: a sound obtaining section that fetches a subject's sound signal; a frequency converting section that converts the sound signal into a frequency spectrum; an auto-correlation section that determines an auto-correlation waveform while shifting the frequency spectrum on a frequency axis; and a pitch detecting section that performs regression analysis on a distribution of the order of occurrence of multiple extremal values of the auto-correlation waveform, and the occurrence frequency representing an amount of a shifted frequency indicative of the occurrence locations of the extremal values, and determines the pitch frequency from the slope of the regression line.
JP-A-2009-294647 (Patent Literature 4) discloses a behavior analysis device and a call center system. This publication provides a sound input section that accepts a subject's speech sound signal; an arithmetic section that calculates a plurality of sound feature amounts from the sound signal; a storage section that stores an emotion table in which a plurality of subject's emotions are associated with a plurality of sound feature amounts, and a behavior table in which the occurrence patterns of a plurality of emotions are associated with subject's behaviors; an analyzing section that uses the emotion table and the behavior table to analyze subject's emotions based on the plurality of sound feature amounts, and predict a subject's behavior; and a display section that outputs and displays the result of the subject's behavior prediction made by the analyzing section.
JP-A-2012-449 (Patent Literature 5) discloses a behavior analysis method and a behavior analysis device. This publication provides an input procedure by which a subject's speech is obtained as a sound signal; an arithmetic procedure by which a plurality of sound feature amounts is calculated from the sound signal; and a predicting procedure by which the physiological activity and the emotion state of the subject are analyzed to predict a subject's behavior by using the plurality of sound feature amounts.
Of particular importance in providing a conservation that a user finds pleasant is how to end a conversation. However, conversations are ended rather unilaterally, or messages to end conversations are not necessarily available in the first place in the sound output devices of related art.